The research methodology refers to the way and manner data is collected for research projects. It includes all the steps from planning and designing the study, to determining and constructing the survey instrument, culminating in sampling and data collection.
This study is an exploratory one, attempting to understand the phenomena of care as given to HIV infected children on treatment and to assess what factors (both medical and non-medical) best influence outcomes. For this reason, qualitative research was considered to be the most useful because it attempts to describe and understand their experiences, beliefs and actions as it impacts on the quality of care provided to children. As this is a non-experimental research, a survey method was adopted. From a large sample (the total number of caregivers of infected children in the Hospital, a target population of about 30 respondents was selected. This selected study sample was achieved by simple random sampling [see section 3.3 below].
As already stated, the research is a survey attempting to gauge the Knowledge, Attitudes and Practices (KAP) of HIV/AIDS caregivers, the ultimate objective of which is to suggest areas for possible improvement in HIV care. Following the relevant Ethics Committee and Institutional approval to conduct the research, the steps involved in the research methodology involved
planning and designing the survey study; survey instrument construction; and sampling and data collection.
3.1. PLANNING AND DESIGNING THE STUDY
The primary factors considered were the time frame for completing the research (the research was meant to be completed over a few weeks/months); the sensitive nature of CGs disclosing details about their HIV positive children/wards and the need to get a heterogeneous group of respondents reflective of the total population. As such, a cross-sectional design was adopted as it fulfilled all the aforementioned criteria. It has been put forth that a cross-sectional study involves identifying representative samples of individuals that differ on some characteristics such as age,
gender, ethnic group, or religion, and measuring these different samples of individuals on the same variable or variable(s) at one point in time (Christensen et al., 2011: 50).
3.2. THE SURVEY INSTRUMENT
The survey instrument used was a questionnaire, which was answered by the research participants. The questionnaire consisted of 35 self-administered and anonymous questions in simplified English, with a Sesotho version for the non-English speaking respondents. This language preference determination was done following consultations with the nurses regarding what major languages were spoken in the clinic. The questions were a mixture of closed and open-ended questions, developed according to experience regarding issues encountered in medical practise, other appropriate and prior research on CGs, and the South African National Department of Health Paediatric HIV Care Manual (2010) with emphasis on the relevant sections on CGs. Areas covered in the questionnaires included:
Demographic characteristics of caregiver; Knowledge about HIV transmission in children;
Knowledge on appropriate nutrition and growth assessment for children with HIV; Assessment of knowledge of co-morbid diseases in HIV-positive children;
Antiretroviral therapy (ART) adherence;
Availability of HIV counselling and support structures; HIV status disclosure to children.
[See Annexures A and B for the respective English and Sesotho versions of the questionnaire.]
3.3. SAMPLE SIZE
Regarding the sample size, the first consideration was recognizing that all the CGs could not be studied, hence the primary consideration was in choosing as accurately representative sample size as possible. The target size was 30 respondents and random sampling (more specifically, simple random sampling) was employed. It has been maintained that random sampling ensures
an accurate depiction of the general population (Christensen et al., 2011). With simple random sampling as the specific method employed, the end result was a method with not only the least sampling error, but one which is most representative of all the total respondents.
A pilot study was conducted on 4 caregivers in an attempt to test the ease of use of the questionnaire so any potential problems could be identified. The entire survey was administered to the respondents (4) and their feedback was used as go-ahead for the survey proper. From this pilot study, the response time was identified to be a mean of 25 minutes each and no complaints were raised by the respondents.
Prior to the process of data collection from the sample population, the purpose of the research was explained to all the CGs who presented at the day clinic on the chosen day, with an opportunity for questions or clarifications. The respondents’ right to participate or decline at any time (without coercion) was also explained and consent forms were provided and their importance elaborated. Respondents were also informed on the language choices (Sesotho and English), and all were assured of confidentiality at all stages of the research, and encouraged to seek clarity at any stage in the process. They were also informed that the study was purely for academic purposes; hence no monetary compensation for participants would be available.
All the CGs of under 5 children (on ART), presenting daily in the out-patient unit of the HIV/AIDS clinic of Kalafong Hospital were assigned numbers, which were folded and placed in a box, from which 30 were blindly and randomly drawn. The ease of choosing this sample from the population was made easy by the fact that Kalafong Hospital has a designated day clinic for under-5 children only. The selected CGs were assessed via a group-administered questionnaire method. This approach involved handing out questionnaires in a group to the selected CGs, so that they were subsequently quickly and efficiently completed. The response time was approximately twenty five minutes each (as evident by a prior pilot study with four randomly chosen CGs from same centre).
Lastly, confidentiality for the respondents was further ensured by providing separate boxes where respondents were asked to place their completed questionnaires in one, and their signed and filled consent forms in the other. This ensured that no consent form was attached to the
sample instrument, so that the information remained completely anonymous and unlinked to any one respondent in particular.
3.4. LIMITATIONS OF METHODOLOGY
This sub-section attempts to discuss some limitations in the methods employed and introduced in this chapter. As such the limitations of the employed methodology will be addressed as follows:
Limitations involving the sample size;
Limitations involving cross-sectional surveys; Limitations involving the use of questionnaires.
3.4.1. LIMITATIONS INVOLVING THE SAMPLE SIZE
Regarding the sample size, the population size (n=30) does not allow for generalizations to be made about caregivers’ KAP in Pretoria. Although the study highlighted certain deficiencies regarding the knowledge and perceptions of certain caregivers studied, the sample size does not allow for statistically significant conclusions to be made. Regarding the sample size in random sampling, Christensen et al., (2011:160) explain that “larger sample sizes make it less likely that you will miss an effect or relationship that is present in your population”. A larger sample size, they postulate, is indicated when the population is heterogeneous (the study attempted to analyze responses from a wide variety of CGs) and the response rate is likely to be low (see Limitations above).
“When simply describing the characteristics of a single group of subjects, the larger the study, the more reliable the results….the main results should have 95% confidence intervals (CI), and the width of this depends on the sample size: large studies produce narrow intervals, and therefore, more precise results” (Hackshaw, 2008, para 2). A CI, for emphasis refers to “a range of numbers inferred from the sample that has a certain probability or chance of including the true population value” (Christensen et al, 2011: 428). A 95% CI invariably has a 95% probability that the population parameter is included.
In summary, although small studies make for quick and easy collection and analysis of data, especially when data is collected in one or a few centres, the main issue with small studies is in the interpretation of results. This is illustrated with a simple equation:
Small studies (result in a large standard error) = Wide 95% CI (imprecise estimate of the effect) = No firm conclusions.
Finally, it is worth pointing out that there exists no rule regarding sample size determination. Despite not usually yielding reliable data, research with small sizes can and should form the template for designing larger studies, and may be used to identify other areas of further research.
3.4.2. LIMITATIONS OF CROSS-SECTIONAL STUDIES
Although easy to design and implement, cross-sectional studies are limited in use because they:
Cannot determine causal relationships. They can only explain correlations between variables. Hence, they have limited in application.
They may require very large sample sizes to be of maximum utility.
They only measure existing (past and present) events. They are of limited use in new or projected events.
3.4.3. LIMITATIONS OF QUESTIONNAIRES USE
Questionnaires as used in this study are limited in the following ways:
Results are usually biased, as participation is voluntary.
The forced-response choice questions do not allow for freedom of expression by respondents.
The response rate is usually low.
Having outlined the methodology that was employed, along with its pros and cons, I shall now undertake a detailed look of what the results revealed in respect of caregivers of children at Kafalong Hospital, Pretoria.